Fast Fourier transform (FFT) is a key algorithm for the real-time tracking and processing of the space-borne ocean radar altimeter. In order to put the complex data into three operations in sequence as multiplying the window, performing FFT, and computing module-square, and also meeting the requirements of high-speed digital signal processing, a novel scheme for implementing FFT with digital signal processing (DSP) C6701 is proposed in this paper. Firstly, Radix-4 algorithm of decimation in time (DIT) is adopted. Then, the improved bit-reverse algorithm is utilized for decreasing the calculation quantum. Pipeline and paralleled pattern are also considered conjointly. Based on these improvements, real-time and precision are all ensured, and the balance of calculation, communication and storage is also well weighted. Finally, a novel approach of flash programming by gel file and second boot loader is adopted, and by using this method, the format conversion of COFF file to hex file is not needed. Compared with MATLAB 6.5, the simulation and system application results all demonstrate the feasibility and effectiveness of this proposed scheme. Furthermore, the design can not only helps following real-time tracking on echo signal of Chinese national HY-2 satellite altimeter, but also provides beneficial engineering experience for FFT implementation in the other embedded systems of DSP.